How to use the Daily Briefing Compiler agent skill (briefing)

What the briefing skill makes the agent do

An agent skill is a folder containing a SKILL.md file that gives an AI agent specific instructions for one task. This skill covers the daily briefing. Before writing anything, the agent runs several read-only lookups against GBrain: salient pages, statistical anomalies, stored personal preferences, recent changes and pending follow-ups.

The finished briefing is organized into sections such as meetings today, active deals, action items, recent changes and people in play. Every fact carries an inline citation in the form [Source: slug, updated DATE], so you can trace any claim back to its page. The skill is designed to be read-only, so it does not create or modify brain pages unless you explicitly ask it to.

• Resolves meeting attendees against your brain and flags people with no record instead of guessing
• Lists active deals with deadlines in the next 7 days
• Surfaces time-sensitive threads with deadlines in the next 48 hours and overdue follow-ups
• Highlights stale pages relevant to today's meetings, not every stale page

Install the briefing skill

You can install the skill in one of two ways. The first is manual: download the ZIP from this page and unzip it into your user skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill limited to that project.

The second is through the Claude apps: upload the ZIP under Settings → Capabilities → Skills.

The package contains a single file, SKILL.md. The skill also refers to other files, skills/_brain-filing-rules.md, skills/google-loops/SKILL.md and conventions/brain-first.md, which the source does not include in this package, so check that they are available if you need the back-linking or open-loops steps.

• Manual install: unzip into ~/.claude/skills/ (all projects) or .claude/skills/ (one project)
• App install: Settings → Capabilities → Skills, then upload the ZIP

Requirements before you trigger it

The skill depends on the GBrain command-line tool (gbrain) and on a knowledge base it can query. Without a GBrain installation, the commands the skill runs will fail, so it will not work out of the box.

Your knowledge base should contain person and deal pages. The open-loops step also needs Google sources synced: the waiting command refuses to run on stale Google sources (no successful sync in 24 hours) and names the fix, which is to run the sync it names and retry.

• GBrain command-line tool (gbrain), with commands such as salience, anomalies, recall, whois, get, query, list, waiting and dream
• A populated GBrain knowledge base with person and deal pages
• Google sources synced for the open-loops step
• Network access, since the skill uses the network

Trigger the skill with natural requests

You do not need a special command. Ask the agent in plain language using one of the trigger phrases the skill lists. These are the phrases in its metadata and when_to_use field.

The skill is meant to run in the morning or whenever you want to check on your day, so a request at the start of the day produces the most useful output.

• "Daily briefing"
• "Morning briefing"
• "What's happening today"
• "Brain pulse"
• "Pre-briefing pull"

What the agent runs before writing

The skill asks the agent to run four read-only context pulls before composing the briefing. Each one feeds a different part of the output.

Salience scan (gbrain salience --days 7) ranks pages by emotional weight and recent activity. Anomaly detection (gbrain anomalies) flags cohorts whose activity broke from their normal cadence. Personal recall (gbrain recall) checks stored preferences and prior commitments so the briefing does not contradict them. The hot memory pulse then runs recall with --since-last-run, --supersessions, --pending and --rollup, and the open-loops step runs gbrain waiting --json.

• Anomalies default to today against a 30-day baseline; use --lookback-days N to widen it or --sigma 2 to lower the threshold
• The --since-last-run flag advances a cursor file at ~/.gbrain/recall-cursors/<source>.json so the next briefing continues where this one stopped

Read an example briefing and its citations

The skill defines the output format below. Each entry links back to a brain page, and the citation shows how current that page is. When a person has no page, the briefing says so and suggests enrichment instead of inventing details.

Citations look like this: "Jane is CTO of Acme [Source: people/jane-doe, updated 2026-04-01]." If an attendee has no page, expect a note such as "No brain page for alice-example -- consider enrichment."

• Meetings today: time, name, participants with slugs and key context
• Active deals: status and deadline, plus latest timeline entry
• Action items: item, due date and related slug
• Recent changes (24h) and people in play

Safety notes: scripts and network

This skill runs scripts and commands and reads personal facts and preferences from your knowledge base. Review what it runs before using it on sensitive data. The source notes that it uses the network. The skill is a poor fit for cron or unattended runs with sensitive data unless you have reviewed its output and commands first.

The skill also states that it does not check live calendars or email on its own. It only reads what is already stored in GBrain, so calendar and email data must reach the knowledge base through your own setup.

• Review the commands in the skill before running it unattended
• Be cautious with cron jobs; the source notes cron does not start in your repo-root directory, so pass --source <slug> or set GBRAIN_SOURCE explicitly
• The skill does not check live calendars or email by itself